Management-induced changes in soil organic carbon  on global croplands

Karstens, Kristine; Bodirsky, Benjamin Leon; Dietrich, Jan Philipp; Dondini, Marta; Heinke, Jens; Kuhnert, Matthias; Müller, Christoph; Rolinski, Susanne; Smith, Pete; Weindl, Isabelle; Lotze-Campen, Hermann; Popp, Alexander

doi:https://doi.org/10.5194/bg-19-5125-2022

Articles | Volume 19, issue 21

https://doi.org/10.5194/bg-19-5125-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/bg-19-5125-2022

© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 19, issue 21

Research article

|

10 Nov 2022

Research article |

| 10 Nov 2022

Management-induced changes in soil organic carbon on global croplands

Kristine Karstens, Benjamin Leon Bodirsky, Jan Philipp Dietrich, Marta Dondini, Jens Heinke, Matthias Kuhnert, Christoph Müller, Susanne Rolinski, Pete Smith, Isabelle Weindl, Hermann Lotze-Campen, and Alexander Popp

Download

Final revised paper (published on 10 Nov 2022)
Preprint (discussion started on 22 Dec 2020)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'review of BG-2020-468', Jonathan Sanderman, 10 Jan 2021
- AC1: 'Reply on RC1', Kristine Karstens, 26 Jan 2021
- AC3: 'Full reply on RC1', Kristine Karstens, 08 Jun 2021
RC2: 'Referee Comments', Anonymous Referee #2, 16 Mar 2021
- AC2: 'Reply on RC2', Kristine Karstens, 08 Jun 2021

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Reconsider after major revisions (11 Jun 2021) by Sönke Zaehle

AR by Kristine Karstens on behalf of the Authors (05 Mar 2022) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (20 Apr 2022) by Sönke Zaehle

RR by Jonathan Sanderman (21 May 2022)

Suggestions for revision or reasons for rejection

The authors have thoughtfully addressed all reviewer comments and have done a thorough job in revising and improving the manuscript. With fresh eyes, I found the manuscript to be extremely clear and insightful, and I believe it will have a big impact on our field. I have only a few comments and suggestions for further discussion of your findings:

L365 - The IPCC2006 values for the tropics were highly criticized as being way too high. The 2019 refinements incorporated >10x as much data. The fact that your results are showing large losses in tropics is problematic and worth some discussion (see additional comment below). Note IPCC factors are derived under the assumption that there is a linear change between steady states over 20 years. Are Fscf factors consistent with this assumption? If not, this might not be a fair comparison.

L378 – This section can go for more than one sentence. Can you describe the results? Goodness of fit for natural veg and cropland (add R2 values to Fig 7)? Individually the results look fairly poor.

Fig 6 - There is a big difference between soilgrids and soilgrids2.0 - in v2.0, they explicitly excluded litter (i.e. organic) horizons on top of dominantly mineral soils. I think this makes your cropland results more comparable to v2.0 but for global stocks, the original soilgrids product may be more reasonable particularly for northern soils where thick litter layers dominate SOC stocks.

L383 - It seems hardly surprising to find good correlation between model results since both models have similar structure albeit with different levels of complexity

L407 – Thinking about structural uncertainly, how would you reconcile your findings with that of the CMIP6-LUMP (Ito et al 2020 ERL - https://iopscience.iop.org/article/10.1088/1748-9326/abc912) which found strong increases in SOC? I think there are strong CO2 feedbacks in those models.

L425 - I disagree that input uncertainty is the largest source of error. Structural uncertainty is likely much larger -- look at inter-model comparison studies such as CMIP5 (Todd-Brown publications) and others that actually include novel model structures (i.e. microbially-explicit model classes)... structural uncertainty leads to differences in +/- 100s of GtC. Wieder et al (https://doi.org/10.1111/gcb.13979) also demonstrated this with a testbed of constant input data.

L455 - again, model to model comparison doesn't mean much if the model structure is similar. The comparison to PNV point data (with the caveat that a large grid cell and a point measurement are hard to reconcile) does not look nearly as good.

L473 – This is not quite accurate. Erosion moves a lot of SOC into aquatic reservoirs, not just to somewhere else on the land, and the 'dynamic' replacement (i.e., high net sink strength on eroding surfaces) of that carbon is what is leading to the offset in losses.

L487 - Another thought on the particularly high losses in the central african basin region -- there is a lot of swidden agriculture practiced here - several years of cropping followed by 10 or more years of fallow to allow soil fertility (and presumably SOC) to rebuild. Is this type of cropping pattern represented in your management data?

L536 – It is not surprising that tillage didn’t have much of an impact. In line with observations, Century model does not have a strong tillage feedback. Dangal et al 2022 (JAMES - https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2021MS002622) showed shifting to no-till had pretty minimal effects of SOC when running DayCent across parts of the US.

Hide

RR by Anonymous Referee #3 (23 May 2022)

Suggestions for revision or reasons for rejection

This paper is a convincing step forward for global cropland modelling of soil carbon, based on land use and agricultural management data that are dynamic in time going back to 1975, and a model framework that is reduced in complexity in order to allow global modelling. The paper is rather long but of great value for the soil carbon and modelling community. The authors clearly state the limitations of the approach and involve several model evaluation steps that are of great help to see the limitations but also the robustness of the results. The paper is written well. At some points it could be improved in order to make it better understandable and easier to read. In particular a clear differentiation between land use change effects and agricultural management effects within croplands would be important. A topic that is not fully explored is the effect of climate change vs. anthropogenic direct impact on global soil carbon dynamics (but this maybe topic for a next paper).

Detailed remarks:

l. 1-2: The first sentences of the abstract are too general and has been repeated too often – please remove.

l. 17: The carbon pool in the lithosphere (including fossil carbon) is much larger than the soil C pool.

l. 41: N-deposition does not play a role in agricultural systems that are fertilized or in areas with no N-deposition. Also, CO2 fertilisation is of minor importance compared to agricultural management including breeding.

L, 46: What are stylized scenarios? And in l. 54: stylized future management? Please explain or rephrase.

l. 56: Please explain why no net sink is possible? In the next paragraph you write about soil C sequestration. The reader might ask why if it is not possible.

l. 106: How did you treat grasslands that are no pastures (mowed meadows e.g.)?

Fig 1: I am not sure if this figure is required to understand the paper. However, please check the values for SOC stocks at the y-axis. They are too low for 0-30 cm SOC stocks for most regions of the world.

l. 169: The assumption that agricultural management from 1510 to 1965 was the same (the same as in 1965) is an assumption with high uncertainty. You might need to explain how important or not important this assumption is for your results.

Chapter 2.2: This consistency check with the IPCC Tier 1 approach is very useful and good.

The start of the results section (l. 304) with two references from the same authors may give the impression that results are published in other papers. Please indicate here that you refer to supplementary studies supporting the actual study with data and code.

Fig 2b) The choice of the colours is suboptimal since the different green cannot be distinguished by eye. Please use more contrasting and different colours. The Fig 2b-d give the impression that most land on earth is cropland even though only around 10% of land surface is cropland.

l. 339: This sentence is not clear to me. It would be better to provide sums of the global input to cropland soils and not to the agricultural system (with undefined system borders). It would be also useful to include terms such as net primary production (NPP) here (also in Fig 4). Also, the term human appropriated fraction of NPP (HANPP) maybe useful also to compare with in the discussion. Derived from the data you stated that 2463 Mt C entered global croplands each year (fig 4). The fraction of manure (16%) is unexpected high since manure can only be transferred to croplands if it is collected (mostly in stables), which is not common in many parts of the world. How much of this manure is feedstock grazing on croplands?

The fraction of above ground biomass C input to the soil (55%) looks to me is quite high, see e.g. Bolinder et al. providing lower fractions. The root fraction should also include rhizodeposits, which might be not considered in your study.

l. 306: This chapter gives a general overview on agricultural land use and management effect on SOC. The impact of land use changes from natural vegetation to croplands on SOC is well known and thus the maps shown here are in many regions of the world in line with the global cropland maps. The new aspect of this study is the agricultural management within croplands. It would be interesting to show maps for agricultural management effects, e.g. a standard or a worst-case scenario vs. the real data scenario. This is part of the next subchapter (3.3.). However, there it is not showed spatially explicit (with maps).

Tab. 3 is not easy to read. For the IPCC values you display low medium and high values. For your modelled data, only an average (I guess medium) value are shown. Would it be possible to also provide low and high values derived from your models for each climate region? It might be also helpful to convert this table into a figure. More important, the effect of agricultural management is not visible in this data set since the IPCC values are developed for land use changes. Again, it would be important to distinguish between land use change and land management effects.

Fig. 6 looks a bit strange with SOC on the x-axis instead of the y-axis. I would also like to see total stocks for croplands and for all other land use types separately.

Chapter 3.4.3: There are no results reported in this chapter for the point data comparison. Moreover, it is questionable if point based data are useful for model validation since SOC stocks can vary at field and local scale considerable and thus a very high number of point data (I would recommend >5000 points globally) of high quality (including bulk density measurements) are required for such an exercise. In the cited database of Sandermann et al. includes less than 300 sites. I think the comparison with the SoilGrid2.0 data is sufficient.

l. 399: There might be a direct link between underestimated SOC and overestimated SOC stock increases and vice versa since models are sensitive with their modeled SOC trend to the initial SOC stock. Thus, it needs to be carefully checked and discussed whether underestimation or overestimation of SOC stocks are the reason for the predicted SOC trends. For example, the strong increase in SOC in arid regions might also be a results of an underestimation of initial SOC stocks.

l. 500: Its not only tillage that can affect subsoil SOC but all land use and land management options can affect SOC in the subsoil below about 30 cm depth. Powslon et al 2014 found not significant tillage effect below 25 cm depth. This might be due to the low sample size of 43 sites. However, there is no evidence that tillage effects subsoil SOC stronger than other agricultural management or land use change.

l. 528: Please provide more detailed data either here or in the results section how increasing yields affected SOC stocks in the past.

l. 544. It would be more helpful to relate and compare the estimated SOC loss with deforestation in Gt/a to the estimated total land use change induced emissions of about 2.5 Gt/a.

l. 549: This sentence (“one fifth of total annual C sequestration by crops is lost through soils (0.8 GtC per year”) is not clear. Please be very careful with the term "C sequestration" throughout the manuscript. C sequestration is the removal and long-term storage of CO2 from the atmosphere. I guess you are referring to the plant photosynthesis flux. The figure 4 is rather complex and may need further description in order to make it understandable also here in your discussion.

Hide

ED: Publish subject to minor revisions (review by editor) (25 May 2022) by Sönke Zaehle

AR by Kristine Karstens on behalf of the Authors (20 Jul 2022) Author's response Author's tracked changes

ED: Publish as is (04 Aug 2022) by Sönke Zaehle

AR by Kristine Karstens on behalf of the Authors (15 Aug 2022) Manuscript

Short summary

Soil organic carbon (SOC) has been depleted by anthropogenic land cover change and agricultural management. While SOC models often simulate detailed biochemical processes, the management decisions are still little investigated at the global scale. We estimate that soils have lost around 26 GtC relative to a counterfactual natural state in 1975. Yet, since 1975, SOC has been increasing again by 4 GtC due to a higher productivity, recycling of crop residues and manure, and no-tillage practices.